Self-supervised learning for classifying paranasal anomalies in the maxillary sinus.

PURPOSE: Paranasal anomalies, frequently identified in routine radiological screenings, exhibit diverse morphological characteristics. Due to the diversity of anomalies, supervised learning methods require large labelled dataset exhibiting diverse anomaly morphology. Self-supervised learning (SSL) can be used to learn representations from unlabelled data. However, there are no SSL methods designed for the downstream task of classifying paranasal anomalies in the maxillary sinus (MS). METHODS: Our approach uses a 3D convolutional autoencoder (CAE) trained in an unsupervised anomaly detection (UAD) framework. Initially, we train the 3D CAE to reduce reconstruction errors when reconstructing normal maxillary sinus (MS) image. Then, this CAE is applied to an unlabelled dataset to generate coarse anomaly locations by creating residual MS images. Following this, a 3D convolutional neural network (CNN) reconstructs these residual images, which forms our SSL task. Lastly, we fine-tune the encoder part of the 3D CNN on a labelled dataset of normal and anomalous MS images. RESULTS: The proposed SSL technique exhibits superior performance compared to existing generic self-supervised methods, especially in scenarios with limited annotated data. When trained on just 10% of the annotated dataset, our method achieves an area under the precision-recall curve (AUPRC) of 0.79 for the downstream classification task. This performance surpasses other methods, with BYOL attaining an AUPRC of 0.75, SimSiam at 0.74, SimCLR at 0.73 and masked autoencoding using SparK at 0.75. CONCLUSION: A self-supervised learning approach that inherently focuses on localizing paranasal anomalies proves to be advantageous, particularly when the subsequent task involves differentiating normal from anomalous maxillary sinuses. Access our code at https://github.com/mtec-tuhh/self-supervised-paranasal-anomaly .

Computer-Aided Diagnosis of Maxillary Sinus Anomalies: Validation and Clinical Correlation.

OBJECTIVE: Computer aided diagnostics (CAD) systems can automate the differentiation of maxillary sinus (MS) with and without opacification, simplifying the typically laborious process and aiding in clinical insight discovery within large cohorts. METHODS: This study uses Hamburg City Health Study (HCHS) a large, prospective, long-term, population-based cohort study of participants between 45 and 74 years of age. We develop a CAD system using an ensemble of 3D Convolutional Neural Network (CNN) to analyze cranial MRIs, distinguishing MS with opacifications (polyps, cysts, mucosal thickening) from MS without opacifications. The system is used to find correlations of participants with and without MS opacifications with clinical data (smoking, alcohol, BMI, asthma, bronchitis, sex, age, leukocyte count, C-reactive protein, allergies). RESULTS: The evaluation metrics of CAD system (Area Under Receiver Operator Characteristic: 0.95, sensitivity: 0.85, specificity: 0.90) demonstrated the effectiveness of our approach. MS with opacification group exhibited higher alcohol consumption, higher BMI, higher incidence of intrinsic asthma and extrinsic asthma. Male sex had higher prevalence of MS opacifications. Participants with MS opacifications had higher incidence of hay fever and house dust allergy but lower incidence of bee/wasp venom allergy. CONCLUSION: The study demonstrates a 3D CNN's ability to distinguish MS with and without opacifications, improving automated diagnosis and aiding in correlating clinical data in population studies. LEVEL OF EVIDENCE: 3 Laryngoscope, 2024.

Nodule detection and generation on chest X-rays: NODE21 Challenge.

Pulmonary nodules may be an early manifestation of lung cancer, the leading cause of cancer-related deaths among both men and women. Numerous studies have established that deep learning methods can yield high-performance levels in the detection of lung nodules in chest X-rays. However, the lack of gold-standard public datasets slows down the progression of the research and prevents benchmarking of methods for this task. To address this, we organized a public research challenge, NODE21, aimed at the detection and generation of lung nodules in chest X-rays. While the detection track assesses state-of-the-art nodule detection systems, the generation track determines the utility of nodule generation algorithms to augment training data and hence improve the performance of the detection systems. This paper summarizes the results of the NODE21 challenge and performs extensive additional experiments to examine the impact of the synthetically generated nodule training images on the detection algorithm performance.

Robust stochastic optimization of needle configurations for robotic HDR prostate brachytherapy.

BACKGROUND: Ideally, inverse planning for HDR brachytherapy (BT) should include the pose of the needles which define the trajectory of the source. This would be particularly interesting when considering the additional freedom and accuracy in needle pose which robotic needle placement enables. However, needle insertion typically leads to tissue deformation, resulting in uncertainty regarding the actual pose of the needles with respect to the tissue. PURPOSE: To efficiently address uncertainty during inverse planning for HDR BT in order to robustly optimize the pose of the needles before insertion, that is, to facilitate path planning for robotic needle placement. METHODS: We use a form of stochastic linear programming to model the inverse treatment planning problem. To account for uncertainty, we consider random tissue displacements at the needle tip to simulate tissue deformation. Conventionally for stochastic linear programming, each simulated deformation is reflected by an addition to the linear programming problem which increases problem size and computational complexity substantially and leads to impractical runtime. We propose two efficient approaches for stochastic linear programming. First, we consider averaging dose coefficients to reduce the problem size. Second, we study weighting of the slack variables of an adjusted linear problem to approximate the full stochastic linear program. We compare different approaches to optimize the needle configurations and evaluate their robustness with respect to different amounts of tissue deformation. RESULTS: Our results illustrate that stochastic planning can improve the robustness of the treatment with respect to deformation. The proposed approaches approximating stochastic linear programming better conform to the tissue deformation compared to conventional linear programming. They show good correlation with the plans computed after deformation while reducing the runtime by two orders of magnitude compared to the complete stochastic linear program. Robust optimization of needle configurations takes on average 59.42 s. Skew needle configurations lead to mean coverage improvements compared to parallel needles from 0.39 to 2.94 percentage points, when 8 mm tissue deformation is considered. Considering tissue deformations from 4  to 10 mm during planning with weighted stochastic optimization and skew needles generally results in improved mean coverage from 1.77 to 4.21 percentage points. CONCLUSIONS: We show that efficient stochastic optimization allows selecting needle configurations which are more robust with respect to potentially negative effects of target deformation and displacement on the achievable prescription dose coverage. The approach facilitates robust path planning for robotic needle placement.

Multiple instance ensembling for paranasal anomaly classification in the maxillary sinus.

PURPOSE: Paranasal anomalies are commonly discovered during routine radiological screenings and can present with a wide range of morphological features. This diversity can make it difficult for convolutional neural networks (CNNs) to accurately classify these anomalies, especially when working with limited datasets. Additionally, current approaches to paranasal anomaly classification are constrained to identifying a single anomaly at a time. These challenges necessitate the need for further research and development in this area. METHODS: We investigate the feasibility of using a 3D convolutional neural network (CNN) to classify healthy maxillary sinuses (MS) and MS with polyps or cysts. The task of accurately localizing the relevant MS volume within larger head and neck Magnetic Resonance Imaging (MRI) scans can be difficult, but we develop a strategy which includes the use of a novel sampling technique that not only effectively localizes the relevant MS volume, but also increases the size of the training dataset and improves classification results. Additionally, we employ a Multiple Instance Ensembling (MIE) prediction method to further boost classification performance. RESULTS: With sampling and MIE, we observe that there is consistent improvement in classification performance of all 3D ResNet and 3D DenseNet architecture with an average AUPRC percentage increase of 21.86 ± 11.92% and 4.27 ± 5.04% by sampling and 28.86 ± 12.80% and 9.85 ± 4.02% by sampling and MIE, respectively. CONCLUSION: Sampling and MIE can be effective techniques to improve the generalizability of CNNs for paranasal anomaly classification. We demonstrate the feasibility of classifying anomalies in the MS. We propose a data enlarging strategy through sampling alongside a novel MIE strategy that proves to be beneficial for paranasal anomaly classification in the MS.

Tissue Classification During Needle Insertion Using Self-Supervised Contrastive Learning and Optical Coherence Tomography.

Needle positioning is essential for various medical applications such as epidural anaesthesia. Physicians rely on their instincts while navigating the needle in epidural spaces. Thereby, identifying the tissue structures may be helpful to the physician as they can provide additional feedback in the needle insertion process. To this end, we propose a deep neural network that classifies the tissues from the phase and intensity data of complex OCT signals acquired at the needle tip. We investigate the performance of the deep neural network in a limited labelled dataset scenario and propose a novel contrastive pretraining strategy that learns invariant representation for phase and intensity data. We show that with 10% of the training set, our proposed pretraining strategy helps the model achieve an F1 score of 0.84±0.10 whereas the model achieves an F1 score of 0.60±0.07 without it. Further, we analyse the importance of phase and intensity individually towards tissue classification.

Bayesian Reconstruction Algorithms for Low-Dose Computed Tomography Are Not Yet Suitable in Clinical Context.

Computed tomography (CT) is a widely used examination technique that usually requires a compromise between image quality and radiation exposure. Reconstruction algorithms aim to reduce radiation exposure while maintaining comparable image quality. Recently, unsupervised deep learning methods have been proposed for this purpose. In this study, a promising sparse-view reconstruction method (posterior temperature optimized Bayesian inverse model; POTOBIM) is tested for its clinical applicability. For this study, 17 whole-body CTs of deceased were performed. In addition to POTOBIM, reconstruction was performed using filtered back projection (FBP). An evaluation was conducted by simulating sinograms and comparing the reconstruction with the original CT slice for each case. A quantitative analysis was performed using peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM). The quality was assessed visually using a modified Ludewig's scale. In the qualitative evaluation, POTOBIM was rated worse than the reference images in most cases. A partially equivalent image quality could only be achieved with 80 projections per rotation. Quantitatively, POTOBIM does not seem to benefit from more than 60 projections. Although deep learning methods seem suitable to produce better image quality, the investigated algorithm (POTOBIM) is not yet suitable for clinical routine.

A systematic approach to deep learning-based nodule detection in chest radiographs.

Lung cancer is a serious disease responsible for millions of deaths every year. Early stages of lung cancer can be manifested in pulmonary lung nodules. To assist radiologists in reducing the number of overseen nodules and to increase the detection accuracy in general, automatic detection algorithms have been proposed. Particularly, deep learning methods are promising. However, obtaining clinically relevant results remains challenging. While a variety of approaches have been proposed for general purpose object detection, these are typically evaluated on benchmark data sets. Achieving competitive performance for specific real-world problems like lung nodule detection typically requires careful analysis of the problem at hand and the selection and tuning of suitable deep learning models. We present a systematic comparison of state-of-the-art object detection algorithms for the task of lung nodule detection. In this regard, we address the critical aspect of class imbalance and and demonstrate a data augmentation approach as well as transfer learning to boost performance. We illustrate how this analysis and a combination of multiple architectures results in state-of-the-art performance for lung nodule detection, which is demonstrated by the proposed model winning the detection track of the Node21 competition. The code for our approach is available at https://github.com/FinnBehrendt/node21-submit.

A Miniature Dual-Fiber Probe for Quantitative Optical Coherence Elastography.

OBJECTIVE: Optical coherence elastography (OCE) allows for high resolution analysis of elastic tissue properties. However, due to the limited penetration of light into tissue, miniature probes are required to reach structures inside the body, e.g., vessel walls. Shear wave elastography relates shear wave velocities to quantitative estimates of elasticity. Generally, this is achieved by measuring the runtime of waves between two or multiple points. For miniature probes, optical fibers have been integrated and the runtime between the point of excitation and a single measurement point has been considered. This approach requires precise temporal synchronization and spatial calibration between excitation and imaging. METHODS: We present a miniaturized dual-fiber OCE probe of 1 mm diameter allowing for robust shear wave elastography. Shear wave velocity is estimated between two optics and hence independent of wave propagation between excitation and imaging. We quantify the wave propagation by evaluating either a single or two measurement points. Particularly, we compare both approaches to ultrasound elastography. RESULTS: Our experimental results demonstrate that quantification of local tissue elasticities is feasible. For homogeneous soft tissue phantoms, we obtain mean deviations of 0.15 ms-1 and 0.02 ms-1 for single-fiber and dual-fiber OCE, respectively. In inhomogeneous phantoms, we measure mean deviations of up to 0.54 ms-1 and 0.03 ms-1 for single-fiber and dual-fiber OCE, respectively. CONCLUSION: We present a dual-fiber OCE approach that is much more robust in inhomogeneous tissues. Moreover, we demonstrate the feasibility of elasticity quantification in ex-vivo coronary arteries. SIGNIFICANCE: This study introduces an approach for robust elasticity quantification from within the tissue.

Real-Time Motion Analysis With 4D Deep Learning for Ultrasound-Guided Radiotherapy.

Motion compensation in radiation therapy is a challenging scenario that requires estimating and forecasting motion of tissue structures to deliver the target dose. Ultrasound offers direct imaging of tissue in real-time and is considered for image guidance in radiation therapy. Recently, fast volumetric ultrasound has gained traction, but motion analysis with such high-dimensional data remains difficult. While deep learning could bring many advantages, such as fast data processing and high performance, it remains unclear how to process sequences of hundreds of image volumes efficiently and effectively. We present a 4D deep learning approach for real-time motion estimation and forecasting using long-term 4D ultrasound data. Using motion traces acquired during radiation therapy combined with various tissue types, our results demonstrate that long-term motion estimation can be performed markerless with a tracking error of 0.35±0.2 mm and with an inference time of less than 5 ms. Also, we demonstrate forecasting directly from the image data up to 900 ms into the future. Overall, our findings highlight that 4D deep learning is a promising approach for motion analysis during radiotherapy.

Automated isocenter optimization approach for treatment planning for gyroscopic radiosurgery.

BACKGROUND: Radiosurgery is a well-established treatment for various intracranial tumors. In contrast to other established radiosurgery platforms, the new ZAP-X® allows for self-shielding gyroscopic radiosurgery. Here, treatment beams with variable beam-on times are targeted towards a small number of isocenters. The existing planning framework relies on a heuristic based on random selection or manual selection of isocenters, which often leads to a higher plan quality in clinical practice. PURPOSE: The purpose of this work is to study an improved approach for radiosurgery treatment planning, which automatically selects the isocenter locations for the treatment of brain tumors and diseases in the head and neck area using the new system ZAP-X® . METHODS: We propose a new method to automatically obtain the locations of the isocenters, which are essential in gyroscopic radiosurgery treatment planning. First, an optimal treatment plan is created based on a randomly selected nonisocentric candidate beam set. The intersections of the resulting subset of weighted beams are then clustered to find isocenters. This approach is compared to sphere-packing, random selection, and selection by an expert planner for generating isocenters. We retrospectively evaluate plan quality on 10 acoustic neuroma cases. RESULTS: Isocenters acquired by the method of clustering result in clinically viable plans for all 10 test cases. When using the same number of isocenters, the clustering approach improves coverage on average by 31 percentage points compared to random selection, 15 percentage points compared to sphere packing and 2 percentage points compared to the coverage achieved with the expert selected isocenters. The automatic determination of location and number of isocenters leads, on average, to a coverage of 97 ± 3% with a conformity index of 1.22 ± 0.22, while using 2.46 ± 3.60 fewer isocenters than manually selected. In terms of algorithm performance, all plans were calculated in less than 2 min with an average runtime of 75 ± 25 s. CONCLUSIONS: This study demonstrates the feasibility of an automatic isocenter selection by clustering in the treatment planning process with the ZAP-X® system. Even in complex cases where the existing approaches fail to produce feasible plans, the clustering method generates plans that are comparable to those produced by expert selected isocenters. Therefore, our approach can help reduce the effort and time required for treatment planning in gyroscopic radiosurgery.

Towards fast adaptive replanning by constrained reoptimization for intra-fractional non-periodic motion during robotic SBRT.

BACKGROUND: Periodic and slow target motion is tracked by synchronous motion of the treatment beams in robotic stereotactic body radiation therapy (SBRT). However, spontaneous, non-periodic displacement or drift of the target may completely change the treatment geometry. Simple motion compensation is not sufficient to guarantee the best possible treatment, since relative motion between the target and organs at risk (OARs) can cause substantial deviations of dose in the OARs. This is especially evident when considering the temporally heterogeneous dose delivery by many focused beams which is typical for robotic SBRT. Instead, a reoptimization of the remaining treatment plan after a large target motion during the treatment could potentially reduce the actually delivered dose to OARs and improve target coverage. This reoptimization task, however, is challenging due to time constraints and limited human supervision. PURPOSE: To study the detrimental effect of spontaneous target motion relative to surrounding OARs on the delivered dose distribution and to analyze how intra-fractional constrained replanning could improve motion compensated robotic SBRT of the prostate. METHODS: We solve the inverse planning problem by optimizing a linear program. When considering intra-fractional target motion resulting in a change of geometry, we adapt the linear program to account for the changed dose coefficients and delivered dose. We reduce the problem size by only reweighting beams from the reference treatment plan without motion. For evaluation we simulate target motion and compare our approach for intra-fractional replanning to the conventional compensation by synchronous beam motion. Results are generated retrospectively on data of 50 patients. RESULTS: Our results show that reoptimization can on average retain or improve coverage in case of target motion compared to the reference plan without motion. Compared to the conventional compensation, coverage is improved from 87.83 % to 94.81 % for large target motion. Our approach for reoptimization ensures fixed upper constraints on the dose even after motion, enabling safer intra-fraction adaption, compared to conventional motion compensation where overdosage in OARs can lead to 21.79 % higher maximum dose than planned. With an average reoptimization time of 6 s for 200 reoptimized beams our approach shows promising performance for intra-fractional application. CONCLUSIONS: We show that intra-fractional constrained reoptimization for adaption to target motion can improve coverage compared to the conventional approach of beam translation while ensuring that upper dose constraints on VOIs are not violated.

Osteocyte apoptosis and cellular micropetrosis signify skeletal aging in type 1 diabetes.

Bone fragility is a profound complication of type 1 diabetes mellitus (T1DM), increasing patient morbidity. Within the mineralized bone matrix, osteocytes build a mechanosensitive network that orchestrates bone remodeling; thus, osteocyte viability is crucial for maintaining bone homeostasis. In human cortical bone specimens from individuals with T1DM, we found signs of accelerated osteocyte apoptosis and local mineralization of osteocyte lacunae (micropetrosis) compared with samples from age-matched controls. Such morphological changes were seen in the relatively young osteonal bone matrix on the periosteal side, and micropetrosis coincided with microdamage accumulation, implying that T1DM drives local skeletal aging and thereby impairs the biomechanical competence of the bone tissue. The consequent dysfunction of the osteocyte network hampers bone remodeling and decreases bone repair mechanisms, potentially contributing to the enhanced fracture risk seen in individuals with T1DM. STATEMENT OF SIGNIFICANCE: Type 1 diabetes mellitus (T1DM) is a chronic autoimmune disease that causes hyperglycemia. Increased bone fragility is one of the complications associated with T1DM. Our latest study on T1DM-affected human cortical bone identified the viability of osteocytes, the primary bone cells, as a potentially critical factor in T1DM-bone disease. We linked T1DM with increased osteocyte apoptosis and local accumulation of mineralized lacunar spaces and microdamage. Such structural changes in bone tissue suggest that T1DM speeds up the adverse effects of aging, leading to the premature death of osteocytes and potentially contributing to diabetes-related bone fragility.

Optical force estimation for interactions between tool and soft tissues.

Robotic assistance in minimally invasive surgery offers numerous advantages for both patient and surgeon. However, the lack of force feedback in robotic surgery is a major limitation, and accurately estimating tool-tissue interaction forces remains a challenge. Image-based force estimation offers a promising solution without the need to integrate sensors into surgical tools. In this indirect approach, interaction forces are derived from the observed deformation, with learning-based methods improving accuracy and real-time capability. However, the relationship between deformation and force is determined by the stiffness of the tissue. Consequently, both deformation and local tissue properties must be observed for an approach applicable to heterogeneous tissue. In this work, we use optical coherence tomography, which can combine the detection of tissue deformation with shear wave elastography in a single modality. We present a multi-input deep learning network for processing of local elasticity estimates and volumetric image data. Our results demonstrate that accounting for elastic properties is critical for accurate image-based force estimation across different tissue types and properties. Joint processing of local elasticity information yields the best performance throughout our phantom study. Furthermore, we test our approach on soft tissue samples that were not present during training and show that generalization to other tissue properties is possible.

Robotic Tissue Sampling for Safe Post-Mortem Biopsy in Infectious Corpses.

In pathology and legal medicine, the histopathological and microbiological analysis of tissue samples from infected deceased is a valuable information for developing treatment strategies during a pandemic such as COVID-19. However, a conventional autopsy carries the risk of disease transmission and may be rejected by relatives. We propose minimally invasive biopsy with robot assistance under CT guidance to minimize the risk of disease transmission during tissue sampling and to improve accuracy. A flexible robotic system for biopsy sampling is presented, which is applied to human corpses placed inside protective body bags. An automatic planning and decision system estimates optimal insertion point. Heat maps projected onto the segmented skin visualize the distance and angle of insertions and estimate the minimum cost of a puncture while avoiding bone collisions. Further, we test multiple insertion paths concerning feasibility and collisions. A custom end effector is designed for inserting needles and extracting tissue samples under robotic guidance. Our robotic post-mortem biopsy (RPMB) system is evaluated in a study during the COVID-19 pandemic on 20 corpses and 10 tissue targets, 5 of them being infected with SARS-CoV-2. The mean planning time including robot path planning is 5.72±167s. Mean needle placement accuracy is 7.19± 422mm.

Systematic analysis of volumetric ultrasound parameters for markerless 4D motion tracking.

OBJECTIVES: Motion compensation is an interesting approach to improve treatments of moving structures. For example, target motion can substantially affect dose delivery in radiation therapy, where methods to detect and mitigate the motion are widely used. Recent advances in fast, volumetric ultrasound have rekindled the interest in ultrasound for motion tracking. We present a setup to evaluate ultrasound based motion tracking and we study the effect of imaging rate and motion artifacts on its performance. METHODS: We describe an experimental setup to acquire markerless 4D ultrasound data with precise ground truth from a robot and evaluate different real-world trajectories and system settings toward accurate motion estimation. We analyze motion artifacts in continuously acquired data by comparing to data recorded in a step-and-shoot fashion. Furthermore, we investigate the trade-off between the imaging frequency and resolution. RESULTS: The mean tracking errors show that continuously acquired data leads to similar results as data acquired in a step-and-shoot fashion. We report mean tracking errors up to 2.01 mm and 1.36 mm on the continuous data for the lower and higher resolution, respectively, while step-and-shoot data leads to mean tracking errors of 2.52 mm and 0.98 mm. CONCLUSIONS: We perform a quantitative analysis of different system settings for motion tracking with 4D ultrasound. We can show that precise tracking is feasible and additional motion in continuously acquired data does not impair the tracking. Moreover, the analysis of the frequency resolution trade-off shows that a high imaging resolution is beneficial in ultrasound tracking.

AI-based optimization for US-guided radiation therapy of the prostate.

OBJECTIVES: Fast volumetric ultrasound presents an interesting modality for continuous and real-time intra-fractional target tracking in radiation therapy of lesions in the abdomen. However, the placement of the ultrasound probe close to the target structures leads to blocking some beam directions. METHODS: To handle the combinatorial complexity of searching for the ultrasound-robot pose and the subset of optimal treatment beams, we combine CNN-based candidate beam selection with simulated annealing for setup optimization of the ultrasound robot, and linear optimization for treatment plan optimization into an AI-based approach. For 50 prostate cases previously treated with the CyberKnife, we study setup and treatment plan optimization when including robotic ultrasound guidance. RESULTS: The CNN-based search substantially outperforms previous randomized heuristics, increasing coverage from 93.66 to 97.20% on average. Moreover, in some cases the total MU was also reduced, particularly for smaller target volumes. Results after AI-based optimization are similar for treatment plans with and without beam blocking due to ultrasound guidance. CONCLUSIONS: AI-based optimization allows for fast and effective search for configurations for robotic ultrasound-guided radiation therapy. The negative impact of the ultrasound robot on the plan quality can successfully be mitigated resulting only in minor differences.

Ultrasound Shear Wave Elasticity Imaging With Spatio-Temporal Deep Learning.

Ultrasound shear wave elasticity imaging is a valuable tool for quantifying the elastic properties of tissue. Typically, the shear wave velocity is derived and mapped to an elasticity value, which neglects information such as the shape of the propagating shear wave or push sequence characteristics. We present 3D spatio-temporal CNNs for fast local elasticity estimation from ultrasound data. This approach is based on retrieving elastic properties from shear wave propagation within small local regions. A large training data set is acquired with a robot from homogeneous gelatin phantoms ranging from 17.42 kPa to 126.05 kPa with various push locations. The results show that our approach can estimate elastic properties on a pixelwise basis with a mean absolute error of 5.01(437) kPa. Furthermore, we estimate local elasticity independent of the push location and can even perform accurate estimates inside the push region. For phantoms with embedded inclusions, we report a 53.93% lower MAE (7.50 kPa) and on the background of 85.24% (1.64 kPa) compared to a conventional shear wave method. Overall, our method offers fast local estimations of elastic properties with small spatio-temporal window sizes.

Posterior temperature optimized Bayesian models for inverse problems in medical imaging.

We present Posterior Temperature Optimized Bayesian Inverse Models (POTOBIM), an unsupervised Bayesian approach to inverse problems in medical imaging using mean-field variational inference with a fully tempered posterior. Bayesian methods exhibit useful properties for approaching inverse tasks, such as tomographic reconstruction or image denoising. A suitable prior distribution introduces regularization, which is needed to solve the ill-posed problem and reduces overfitting the data. In practice, however, this often results in a suboptimal posterior temperature, and the full potential of the Bayesian approach is not being exploited. In POTOBIM, we optimize both the parameters of the prior distribution and the posterior temperature with respect to reconstruction accuracy using Bayesian optimization with Gaussian process regression. Our method is extensively evaluated on four different inverse tasks on a variety of modalities with images from public data sets and we demonstrate that an optimized posterior temperature outperforms both non-Bayesian and Bayesian approaches without temperature optimization. The use of an optimized prior distribution and posterior temperature leads to improved accuracy and uncertainty estimation and we show that it is sufficient to find these hyperparameters per task domain. Well-tempered posteriors yield calibrated uncertainty, which increases the reliability in the predictions. Our source code is publicly available at github.com/Cardio-AI/mfvi-dip-mia.

In vivo detection of head and neck tumors by hyperspectral imaging combined with deep learning methods.

Currently, there are no fast and accurate screening methods available for head and neck cancer, the eighth most common tumor entity. For this study, we used hyperspectral imaging, an imaging technique for quantitative and objective surface analysis, combined with deep learning methods for automated tissue classification. As part of a prospective clinical observational study, hyperspectral datasets of laryngeal, hypopharyngeal and oropharyngeal mucosa were recorded in 98 patients before surgery in vivo. We established an automated data interpretation pathway that can classify the tissue into healthy and tumorous using convolutional neural networks with 2D spatial or 3D spatio-spectral convolutions combined with a state-of-the-art Densenet architecture. Using 24 patients for testing, our 3D spatio-spectral Densenet classification method achieves an average accuracy of 81%, a sensitivity of 83% and a specificity of 79%.